Existing Methods of Automated Measurement
Automated measurement of medium to large size objects requires a measuring machine accuracy of 0.05 mm (+/−2 Sigma) and typically 0.025 mm (+/−2 Sigma) or better. ‘Sigma’ means one standard deviation. It is currently carried out in two main ways: (i) a bulky, expensive, conventional Computer Numerically Controlled Coordinate Measuring Machine (CNC CMM) with 3 or more axes; (ii) a rigid structure of static Optical probes that is typically located in a dedicated cell at the end of the automotive production line. With a conventional CMM, the Optical probe moves in a highly controlled way around a static object to produce accurate data. In the second case, both Optical probes and object are static and localised in a calibrated way that permits accurate data. Most conventional CMMs are of either the moving bridge or horizontal arm structures; companies including Zeiss (Germany), Hexagon Brown&Sharpe (Sweden) and LK (U K) produce them. Mechanical touch probes for mounting on conventional CMMs are supplied by companies including Renishaw (UK). Optical probes for mounting on conventional CMMs are supplied by companies including Metris (Belgium). Automatic probe mounts such as the Renishaw Autojoint are repeatable to a high degree of accuracy and are supplied with a rack of probes for automatic probe changing. Rigid structures of static Optical probes are supplied by Perceptron (USA). Both conventional CMMs and rigid structures of static Optical probes have the disadvantages that: they use up cell space on a production line that is typically only used for measurement and not a productive operation, they are usually situated at the end of the line, cannot feed forward data to downstream processes and are expensive and are difficult to justify on a payback basis. In addition, rigid structures of Optical probes are inflexible for rapidly changing models on the production line. Today, efficient production processes using robots that are quicker, better or cheaper than conventional processes but require high accuracy location cannot be deployed on the production line because of the disadvantages of existing high accuracy measurement systems.
Robot Automated Measurement
Since the 1960s, companies have developed heavy robot arms for applications requiring quick cycle times and repeatability. However, due mainly to temperature, wear and vibration problems, they have low accuracy. Robots have been used to carry probes for automated measurement. The robot arms are not accurate enough to meet the demanding requirements of most automated measurement, particularly in the automotive industry. The high repeatability of a robot arm has made ‘quasi-static’ measurement a solution that has received some uptake by the automotive industry. In ‘quasi-static’ measurement, the probe is moved from one position to the next and only takes data when static or moving slowly. Measurement can be by either contact or non-contact probes. Measuring probes on robot arms taking three-dimensional data from the surface of an object whilst moving at typical speeds of 10 mmsec-200 mm/sec (but can be more or less) are not accurate. Companies producing robot arms include Fanuc (Japan) and Kuka (Germany). Perceptron and LMI-Diffracto (USA) offer solutions using robot arms and Optical probes. 3D Scanners and Kuka showed a solution with real-time optical inspection at the Euromold 2001 exhibition in Frankfurt; its accuracy was of the order of 0.5-1 mm. Standard industrial robots thermally grow by around 10 microns per degree Celsius temperature increase per meter of reach; errors in excess of 500 microns can be recorded in production line conditions. LMI-Diffracto have an automotive production line installation comprising four standard industrial robots supplied by Kuka, each carrying an Optical probe, wherein the robots are compensated for thermal growth, potentially reducing the thermal error in production line conditions to below 100 microns. In U.S. Pat. No. 6,078,846 Greer assigned to Perceptron, compensation for robot thermal growth is carried out by measuring a fixed artefact with the Optical Probe. The Optical probes measure whilst the robot is static between movements. Error mapping has improved robot accuracy. There are several approaches including dancing the robot through a program of planned movements whilst measuring it with a photogrammetric system such as that from Krypton (Holland) or Northern Digital (Canada). The measurements are then used to create an error map. Error compensation for load has been carried out by measuring the power used by the servos to automatically calculate the loads on the arm. Even with multiple types of error compensation, accuracies of only 0.2 mm (+/−2 Sigma) have been achieved for robots of the type and reach found in large quantities on automotive production lines. The problem with robot arms carrying scanning probes in which there is relative movement between the probe and the object during scanning is that the systems are not accurate enough to be useful.
Tracking
In U.S. Pat. No. 6,166,811 Long et al, a photogrammetry system for increasing the accuracy of scanning an object is disclosed in which photogrammetry targets affixed to the probe are tracked by a photogrammetry system in real-time. There are many disadvantages to this method. Firstly, a plurality of clear lines of sight need to be maintained between the probe and the photogrammetric cameras. In practice, lines of sight from the photogrammetry cameras to the photogrammetry targets on the probe are often blocked by the programmed robot movement and or the programmed changes in probe orientation necessary to scan the object. This so constrains the applicability of the system as to render it useless for many applications. Secondly, environmental lighting conditions must be maintained at a near ideal state or the accuracy of the photogrammetric system will reduce or the system will cease to function. In practice this is difficult to set up and often conflicts with other lighting requirements for the location. Thirdly, photogrammetric systems often do not have both the resolution and speed necessary for providing sufficient accuracy in this application. Fourthly, the photogrammetric cameras and the Robot must be mounted rigidly relative to each other. This often necessitates a stiff structure of large dimensions to achieve the desired accuracy. The main problem with incorporating photogrammetric technology into a robot measuring system is that the resulting systems are not compact and robust enough to be useful.
Leica Geosystems supply the 6 degrees of freedom Laser Tracker LTD800. It can measure position and orientation over a 35 m range with a single line of sight at up to 1000 measurements per second. Its accuracy is of the order of 50 microns for slow moving targets. Its cost is in excess of US$130,000. Many of its limitations for robot measuring are similar to those of photogrammetry. The main problems with incorporating laser tracker technology into a robot measuring system is that it is expensive, there are limitations to the orientation of the probe being tracked and the resulting systems are not compact and robust enough to be useful.
Robot Controllers and Programming
Controllers for robot arms are well understood by those skilled in the field; a standard reference work is ‘Robot Manipulators, Mathematics Programming and Control’ by Richard P Paul. Adept Technologies (US) supply 6-axis robot controllers starting at US$8,500. There are many products available for the programming of robots that allow motion sequences to be generated off-line and subsequently communicated to the Robot Controller for later execution; one example is EmWorkplace from Tecnomatix (US). In Patent Application GB 2036376A Richter assigned to HA Schlatter AG (Switzerland), programming is achieved by manually guiding a robot by means of a device mounted on the robot that is held by the user and comprises strain gauges that detect the user's intended direction for robot.
Manual CMM Arms
Since the 1970's, companies have been building manually operable CMM arms that have recently achieved a measuring accuracy using a contact probe of between 0.025 mm (+/−2 Sigma) and 0.005 mm (+/−2 Sigma) depending, mainly, on the reach of the Manual CMM Arm. Manual CMM Arms are expected to become more accurate with further development. These Manual CMM Arms are now accurate enough for many measurement requirements and are a growing sector in the measurement marketplace. They have the flexibility of being able to get into areas with difficult access. Manual CMM Arms are acceptably accurate for many applications, but are not automated; they are expensive to operate, particularly since a semi-skilled operator is required; human operators are also subject to human error. Manual CMM Arms are produced by companies including: Cimcore (USA), Faro Technologies (USA), Romer (France), Zett Mess Technik (Germany) and OGP (UK). As examples, U.S. Pat. No. 3,994,798 Eaton, U.S. Pat. No. 5,402,582 Raab assigned to Faro Technologies, U.S. Pat. No. 5,829,148 Eaton and U.S. Pat. No. 6,366,831 Raab assigned to Faro Technologies disclose background information on Manual CMM Arms. The provision of bearings at the joints of Manual CMM Arms is well known and U.S. Patent Application 2002/0087233 Raab assigned to Faro Technologies discloses background information on bearings. The design of Manual CMM arms is typically limited to around 2 metres in reach from the centre of joint 1 to the probe tip because any longer and it requires two operators to use the arm. The longer the Manual CMM arm is, the less accurate it is. In general, for a modular Manual CMM Arm design all other things being equal, the accuracy is at best inversely proportional to the length. In U.S. Pat. No. 6,366,831 Raab, it is disclosed that in the field, Manual CMM Arms typically have an absolute positional accuracy ten or more times that of a robot arm. Some of the factors in robots that cause inaccuracy including joint misalignments are referred to in U.S. Pat. No. 6,366,831. Manual CMM arms such as those manufactured by Faro Technologies and Romer are generally operated by a single person using both hands. Each of the operator's hands provides a different 6 DOF action on the segment of the Manual CMM arm that is gripped by the hand. Some skilled operator's may only need one hand in some applications. A Manual CMM arm is a mechanism that is controlled in a closed-loop fashion wherein the operator closes the loop. Such control is a skilled activity; the operator needs to control 6 or 7 axes of arm freedom in a variety of different spatial layouts, under the effect of gravity, with just two hands. It is often the case that the operator mishandles the Manual CMM arm and part or all of the Manual CMM arm accelerates under gravity until there is a collision or the operator steadies it. It is the case that during data capture, the operator applies variable and occasionally excessive forces and torques on the Manual CMM arm, which reduce the accuracy of the measurement data the Manual CMM arm outputs.
Compensating and Holding Devices
A Manual CMM Arm typically has a compensating device built into the second joint that provides a torque on the upper arm that tends to provide a lifting force on the upper arm to counterbalance it. Compensating devices for manual CMM arms are disclosed in U.S. Pat. No. 6,298,569 Raab et al, U.S. Pat. No. 6,253,458 Raab et al, and U.S. Patent Application 2003/0167647 Raab et al, all assigned to Faro Technologies. This means that the arm is lighter for the operator to lift and is consequently less tiring to use. This also means that more torque is transmitted through the Manual CMM Arm and requires that the Manual CMM Arm must be designed to be heavier than without such a compensating device to achieve a required accuracy. It is standard practice to compensate Robots in order to reduce robot power consumption and the power, size and weight of the motors. In 2003/0167647, a machined spring compensating device can be removed, reversed and replaced to compensate the arm when used in a hanging down orientation; this procedure is inconvenient for the user since it must be carried out in the factory. Some Manual CMM Arms have holding devices to lock one or more axes of the arm in any spatial orientation; such holding devices eliminate the need to lay the arm down between sets of measurements. On the 3000 Series manual CMM arm from Cimcore (USA) there is a sliding peg fixing mounted on the compensating device on Axis 2 (the first orthogonal hinge axis); when the peg slides into a hole, the compensating device on which Axis 2 rests is locked. A pneumatic brake on several axes is disclosed in PCT/EP01/01570 Nietz assigned to Zett Mess Technik GmbH and offered on Axes 1 to 4 of Zett Mess's AMPG-P manual CMM arm product; the pneumatic brakes can be released by radio remote control switch; the pneumatic brakes act on a disk. The pneumatic brakes and disks are mounted directly on the manual CMM arm; they add weight to the manual CMM arm and pass moments through the bearings of the manual CMM arm, thereby reducing its accuracy and usability.
Optical Probes on Manual CMM Arms
Optical probes on Manual CMM Arms were disclosed by Crampton, the inventor of the present invention, in several patent applications including WO9705449. Optical probes for Manual CMM Arms are provided or are being developed by 3D Scanners, Romer, Faro Technologies, Perceptron, Steinbichler (Germany), Pulstec (Japan) and Kreon (France) amongst others. Optical probes are generally mounted offset on the side of the Manual CMM Arm or mounted on the probe end of it. There are three broad types of Optical probe: point, line and area. As yet, a measurement accuracy standard does not exist that defines the way accuracy should be measured for point, line and area Optical probes. The marketplace is in the situation of not being able to perform standard tests to verify accuracy and enable comparison between Optical probe types in a practical way. Optical probes have become accurate, largely because their measuring range is short. In general, Optical probes gather measurement data over a measuring range of the order of 20-400 mm. This is often at a standoff to the end of the Manual CMM Arm. The accuracy of the best Manual CMM Arms combined with the best Optical probes is already better than 0.050 mm (+/−2 Sigma) and can be better than 0.010 mm (+/−2 Sigma) or even 0.002 mm (+/−2 Sigma) for short measuring ranges.
Synchronisation and Interpolation of Optical Probes on Manual CMM Arms
In a system comprising a Manual CMM Arm and an Optical probe, measurements from each are combined to give the output measurement data. As disclosed in WO9705449 by Crampton, the inventor of the present invention, the measurement accuracy of a system comprising a Manual CMM Arm and an Optical probe is increased by synchronising the timing of a measurement from the Manual CMM Arm and a measurement from the Optical probe. As further disclosed in WO9705449, alternatively, the measurement accuracy of a system comprising a Manual CMM Arm and an Optical probe is increased by time-stamping each measurement from the Manual CMM Arm and time-stamping each measurement from the Optical probe and later using a process of interpolation of the two sets of measurements to provide a combined set of measurements. However, occasionally there is a perturbation in the system and one or more measurements from one device or another are lost. In this situation, the later process of interpolation can be complex.
Calibration and Alignment of Robots and Manual CMM Arms
As disclosed in U.S. Pat. No. 5,687,293 Snell, a robot can be calibrated using a reference sphere and a spherically tipped probe on the robot by bringing the spherically tipped probe into contact with the reference sphere a number of times with different robot spatial layouts; a 39-parameter kinematic model for a 6-axis robot embodiment is disclosed. The alignment of Optical probes to Robots is disclosed in U.S. Pat. No. 6,321,137B1 De Smet. A method of manually calibrating a Manual CMM Arm is disclosed in U.S. Pat. No. 5,402,582 Raab assigned to Faro Technologies. Manual CMM Arms are calibrated by the manufacturer before shipping. Some suppliers, including Faro Technologies, enable the user to perform a simple probe calibration each time the probe is changed, whilst the Manual CMM Arm calibration remains the same. OGP UK supply the Polar Manual CMM Arm and permit the user to fully calibrate the Polar arm and probe together in a simple procedure by using a reference artefact with several cones into which the spherical probe of the Polar arm is placed whilst the arm is exercised through a variety of spatial layouts; a 39-parameter kinematic model is used for their 6-axis Polar arm. Full and accurate manual calibration of Manual CMM Arms is a painstaking process in which typically 500 separate points are recorded with a process taking some hours. Each point is subject to human error. Different operators hold Manual CMM Arms in different locations, apply different torques through different grips, applying different patterns of loads and bending moments on the arm resulting in different deflections and end slopes. A Manual CMM Arm that is manually calibrated will perform differently depending on how each operator holds and uses it. A Manual CMM Arm is required that is under repeatable patterns of loads and bending moments however it is held for each spatial orientation. A manual method of calibrating Manual CMM Arms is required that has the same pattern of loads and bending moments that will occur in its use by different operators. An automated method of calibrating Manual CMM Arms is needed to increase the repeatability and accuracy of their calibration, in particular enabling more points to be recorded than is practical or cost effective in current manual processes. The alignment (also known as calibration or qualification) of Optical probes to Manual CMM Arms is disclosed in WO9705449 by Crampton, the inventor of the present invention.
Attachment of Robots and Measuring Devices
As disclosed in U.S. Pat. No. 5,392,384 Tounai et al, the tip of a 6 axis articulated measuring device is attached to the tip of a robot for the purpose of calibrating the robot. As disclosed in U.S. Pat. No. 6,535,794 Raab assigned to Faro Technologies the tip of a 6 axis articulated measuring device is attached to the tip of a robot for the purpose of generating an error map. As disclosed in U.S. Pat. No. 6,519,860 Bieg et al, the tip of a 3 axis articulated measuring device is attached to the tip of a robot or machine for the purpose of measuring the spatial performance of the robot or machine. Neither of these disclosures is used to measure an object. As disclosed in WO 98/27887 Wahrburg, a surgical robot and a multiple joint sensor arm are attached at the base; the multiple joint sensor arm is used manually to make measurements on the patient, a robot program is generated based on those measurements and the robot carries out the surgical intervention. In this disclosure the measurement is not automated. Two items of prior art disclose devices for measuring the position and or orientation of the endpoint of a robot arm subject to deflections from bending and or thermal expansions. As disclosed in U.S. Pat. No. 4,119,212 Flemming, a simple knee joint with a planar goniometer rigidly attached at both ends is used to monitor the location of the end of the moving segment. The device is limited to operation in a plane and no out of plane bending is measured. It is therefore not able to measure position and orientation in a 3D space. As disclosed in U.S. Pat. No. 4,606,696 Slocum, a device for measuring the position and orientation of the end of a robot arm comprises a multitude of measuring links joined by rotary and linear bearings and measuring devices to measure rotational angles and linear movements. As well as being pinned at both endpoints of the robot arm, measuring links are rigidly pinned to the robot arm in at least one intermediate hinge joint. This approach requires 12 accurate rotary and linear measuring devices on a 6-axis robot. The stack-up of errors from the 12 measuring devices call into question whether it could ever be developed into an accurate 3D measuring device for 6-axis robots. A simpler and more robust system is needed that does not require additional rotary and linear measuring devices and their associated error stack-up. Both U.S. Pat. Nos. 4,119,212 and 4,606,696 require rigid attachment of the measuring device at each end of the robot arm. Rigid attachment at the probe end is essential for precisely measuring the position of the end of the robot arm. Rigid attachment at the probe end is neither required nor desirable when the robot arm is used to position a CMM arm. Neither U.S. Pat. No. 4,119,212 nor U.S. Pat. No. 4,606,696 provide means for using calibration information in the devices. Neither do they propose the use of the devices as Coordinate Measuring Machines. Without the use of calibration information, it is questionable whether the devices could be anywhere near as accurate as is required in current applications.
Other Background
As disclosed in PCT/GB01/01590 Gooch, a robot is shown with both an optical probe and a tool mounted at the probe end of the robot; the robot can be used alternately for measuring with the optical probe and performing an operation with the tool; however, to achieve measuring accuracy, an optical tracking system is used that has all the disadvantages previously mentioned. As further disclosed in PCT/GB01/01590 Gooch, a robot may be mobile, for example mounted on rails, to provide access around a large object being measured; this further disclosure also has the disadvantages of optical tracking. A manual marking out system utilising the Faro arm and a robot marking out system utilising an industrial robot from Kuka are disclosed in PCT/GB01/03865 Gooch; these two systems are either accurate or automated but not both. Manual scanning of an object on a rotary table with a non-contact sensor mounted on a Manual CMM Arm is disclosed by Crampton, the inventor of the present invention, in patent application WO9705449. Milling of large objects has been carried out by standard 5 or 6 axis Industrial Robots; the resulting object is not accurate due to the limitations in accuracy of standard industrial robots and typically requires hand finishing where cuts form different orientations result in Milling of large objects is regularly carried out accurately on large 5-axis machining centres such as those manufactured by Mecof spa (Italy) and on large 5-aixs horizontal arm CMMs such as those from Zeiss and LK Tool; the class of objects that may be machined is limited by the Cartesian machine type, for instance a horizontal arm cannot bend around corners. Delcam (UK) provide a software called PowerShape that is capable of generating milling programs for 5 axis and 6 axis industrial robots.
Need for Accuracy
Users demand ever higher accuracy from their Manual CMM Arms. A significant amount of error in Manual CMM Arms is derived from the operator over-stressing the Manual CMM Arm, variability of moments on the arm through different hand grip positions and built-in counterbalances providing moments across bearings. There is a need for a more repeatable Manual CMM Arm in which the loads on the CMM Arm are independent of how it is held and which is significantly more accurate. There is a further need for a more accurate calibration process that is automated to remove human error.
Need for Automation
A Manual CMM Arm with an Optical probe is typically used for many hours at a time. During much of this time, the operator holds the Manual CMM Arm at a distance from him, often in awkward locations. The weight that is supported at a distance can be several kilograms for a long Manual CMM Arm. This is hard work and is tiring for many operators, particularly smaller people; operator fatigue is a common problem and this can lead to illness, incapacitation or injury. Much of the work done with Manual CMM Arms is for unique components that only need to be optically inspected once. Often, the surface being inspected is not immediately accessible and requires temporary gantries to be erected for the operator to climb on so that the arm can be manipulated. The problem with Manual CMM Arms carrying scanning probes in which there is relative movement between the probe and the object during scanning is that, although they are accurate enough, the system is fatiguing to use and can output inaccurate data through operator error or over-stressing of the Manual CMM Arm, because it cannot operate automatically.
Need for Accessibility
The shape of objects to be measured and their accessibility to a probe on a movable member vary from application to application. A CMM that is flexible enough to access a larger range of object shapes has more utility. In practice, it is generally established that articulated arm CMMs comprising a series of preferably 6 or 7 joints separated by rigid segments are more flexible than orthogonal axis configuration CMMs. It is also generally established in the existing state of the art that automated orthogonal axis configuration CMMs are several orders more accurate than automated articulated Robot arms. It is also generally established that automated orthogonal axis configuration CMMs are less suitable than automated articulated arm Robots for locating in a manufacturing environment such as on an assembly line. The problem is that no automated CMM machine is available that is articulated and sufficiently accurate.
Need for Portability
As exhibited by the purchase of the order of around 5,000 portable Manual CMM Arms since they became accurate enough in the mid-1990's, there is a significant demand for portable Manual CMM Arms. There is a corresponding need for a portable Automated CMM Arm, but none exist today.
Need for Robustness
Manual CMM Arms are becoming more accurate and less robust. The existing designs of Manual CMM Arms have the precision measuring system exposed to shocks, moments and abuse in usage and transportation. Existing designs of transportation case are unsophisticated and expose Manual CMM Arms to damage, particularly from shock. There is a need for a robust portable Manual CMM Arm and a transportation case that minimises forces and moments on the Manual CMM Arm from shocks in transportation.